Cluster Analysis and Priority Sorting in Huge Point Clouds for Building Reconstruction

Abstract

Terrestrial laser scanners produce point clouds with a huge number of points within a very limited surrounding. In built-up areas, many of the man-made objects are dominated by planar surfaces. We introduce a RANSAC based preprocessing technique that transforms the irregular point cloud into a set of locally delimited surface patches in order to reduce the amount of data and to achieve a higher level of abstraction. In a second step, the resulting patches are grouped to large planes while ignoring small and irrelevant structures. The approach is tested with a dataset of a builtup area which is described very well needing only a small number of geometric primitives. The grouping emphasizes man-made structures and could be used as a preclassification.

Stereo analysis of high-resolution SAR images for building height estimation in cases of orthogonal aspect directions

“…In this case Aziz and Mertsching (2007) show that inhibition functions from already observed gestalts is a reasonable heuristic. Accordingly, we have used Gaussian spatial inhibition functions, which we developed earlier for grouping of co-planar laser-rangefinder spots in urban terrain (von Hansen et al, 2006). But, to our knowledge in Gestalt theory the step from a probability centered approach like the one of Desolneux (2000) to a real theory of utility incorporating these inhibition principles has not yet been formulated.…”

A Fast and Accurate Segmentation Method for Ordered LiDAR Point Cloud of Large-Scale Scenes

“…The model fitting method exemplified by RANSAC [6] is relatively fast, while fitting the points to certain primitive shapes such as plane, cylinder, and sphere has a limitation for accurate segmentation. The clustering method [7] based on distances to identify clusters cannot be used alone for largescale segmentation since the objects located on the ground will be segmented into the same cluster with the ground. The cluster extraction algorithm [8] combines the RANSAC and the Euclidean clustering method.…”

“…This algorithm detects the locally lowest points and uses them as the seeds to grow other points of ground with slope and elevation. The accuracy of this method is limited since this work is based on aerial systems, and thus, the tolerance of height difference between two points has to be set as 1 m. For unordered input data, the related methods can be further categorized into four groups, namely, region growing [4], tensor voting [5], model fitting [6], and clustering [7]. The region growing method [4] and the tensor voting method [5] both have to investigate the relationship between any a point and its neigh- bor points, which results in a huge complexity and memory usage.…”

Ad Hoc Model Generation Using Multiscale Lidar Data From a Geospatial Database

“…A region growing approach for plane extraction and a subsequent convex hull determination for polygon generation is presented in (Vaskevicius et al, 2007). To speed up the plane extraction, the methods in the following publications separate the points into cells, either grid or octree, and try to extract a plane per cell: A grid approach with RANSAC plane extraction is shown in (Hansen et al, 2006). Two similar solutions using an octree data structure are presented in (Wang and Tseng, 2004) and (Jo et al, 2013).…”